Srsly Risky Biz: Iranian Hacks With Russian Characteristics

Your weekly dose of Seriously Risky Business news is written by Tom Uren and edited by Patrick Gray and Amberleigh Jack. This week’s edition is sponsored by Airlock Digital.

You can hear a podcast discussion of this newsletter by searching for “Risky Business News” in your podcatcher or subscribing via this RSS feed.

French authorities have attributed recent supermarket freezer hacks to cyber units within the Iranian Revolutionary Guard Corp (IRGC). 

These attacks felt very Russian, so we’re surprised.

It’s important to know what exactly has been attributed here. There’s been a string of refrigeration hacks since an initial incident in August, when freezers at US military commissaries suddenly stopped working at at least six different bases.

In October, French retailers were hit. Then came a huge wave of supermarket attacks in early November, affecting more than 115 stores in Belgium, the Netherlands, Sweden, Luxembourg, Ireland and Estonia. 

Finally, French national blood service fridges were hacked last week and their temperature settings raised. Fortunately, monitoring systems alerted staff to the problem and they were able to intervene before any damage was caused. 

The specific incident attributed to Iran by French authorities was the string of attacks that hit French retailers in October, from Sunday 11 to Monday 12. Per Risky Bulletin: 

The hack hit 700 retail stores across France, but also a few butchers, food factories, and even a slaughterhouse. While the Intermarché group was heavily hit, the attacks also crippled freezers at supermarket chains such as Casino, Leclerc, Coopérative U, Metro, and Carrefour.

The hacked freezers were set to super high temperatures that defrosted and spoiled food, but in 15 locations, the freezers overheated and also triggered fires. 

The attacks mainly targeted the web panels of Danfoss refrigeration controllers, which allow store IT teams and employees to control the temperatures of their freezers from centralized panels or over the internet.

Unnamed US intelligence sources told The New York Times they had reached the same conclusion, but only with low confidence. More intelligence could change their assessment.   

That whodunnit confusion is just one reason these sorts of attacks are so hard to deter. Three of the attacks have not yet been attributed and for the one that has, the US is hedging. Exactly how strong is the evidence that Iran is responsible? As already mentioned, if we’d been forced to bet, we would have put our chips on Russia being responsible as the attacks are consistent with its destructive hybrid attacks across Europe in recent years.

Russian cyber operators have masqueraded as Iranian threat actors in the past. Back in 2019 the UK’s NCSC reported that the Russian group Turla had hijacked Iranian tools and malware to conduct espionage across the Middle East. 

It also doesn’t make much sense for Iran to cause chaos across Europe. Iran’s cyber shenanigans earlier this year were firmly focussed on US water infrastructure. In our view, those attacks were well calibrated to achieve US political impact without any significant risk of escalation in its conventional conflict. Whereas Iranian attacks on European refrigeration… just don’t really track. 

The New York Times also reported that Western former security officials were at a loss to explain why Iran would target Europe, given that it is in the middle of a war with the US.

As for the French government, it’s said it’ll summon the Iranian Ambassador and express its strongest condemnation. Additionally, it will talk to European partners about the adoption of new sanctions. A stern talking to and some sanctions! That’ll totally work!

Whether France is right and Iran is responsible, or it turns out that Russia was muddying the attribution waters, deterring either of them will be difficult. Both are involved in serious conventional conflicts and are already sanctioned up the wazoo. 

Whoever is responsible, they have found a rich vein of annoyance to mine. We just hope smart freezers aren’t the thin edge of the wedge.

China Adopts Everything Everywhere All at Once Strategy

Back in August we covered how one Chinese cyber espionage actor, SilkParasite, was using AI to diversify its malware arsenal to make it more difficult to attribute its various hacking campaigns. That makes sense when an important goal is to operate quietly and not get caught. 

By contrast, China’s B-team hackers are using AI to go faster, albeit noisily. 

China has embraced stealthy and noisy operations, depending on its campaign objectives, and recent threat intelligence reports bear this out.

Starting on the noisy side, CrowdStrike released a report last month on a Chinese cyber espionage actor using AI for a pedal-to-the-metal hacking campaign that prioritises speed and scale, stealth be damned. The actor, which CrowdStrike calls Vapor Panda, is using the open-source agentic pentesting framework ARTEX to ramp up its previous hacking capability.

The Google Threat Intelligence Group (GTIG) describes ARTEX as “an autonomous, multi-agent LLM-driven pentesting and offensive command-and-control (C2) platform”. ARTEX can hook into both frontier and open source models and it’s definitely not just China that’s cottoned on to its usefulness. 

When it comes to Vapor Panda, Crowdstrike says ARTEX “pursued multiple attack paths in parallel” including web applications, cloud services, container environments, mail and VPN infrastructure. If any path was unsuccessful it would shift to another alternative approach. A single operator could use ARTEX to run multiple simultaneous engagements, each targeting different organisations, and only jumps in to direct the agents when decisions are required. 

Vapor Panda certainly did plenty of hacking with ARTEX. In the space of 11 days beginning in mid-October the group used it against more than 200 organisations. These, according to CrowdStrike, included “foreign ministries in Latin America, Europe, Asia, and the Middle East [and] semiconductor, defense, and think tank organisations in the US”. They also targeted “telecommunications providers in multiple regions”. 

Of course, using ARTEX to hack quickly is not exactly subtle. GTIG says the harness relies on “rapid, iterative trial-and-error” and that compared to a competent human it is “forensically noisy”. In one Vapor Panda intrusion, ARTEX’s interaction with just a single web server  “generated 240 log events and exhibited errors potentially caused by multiple sub-agents overwriting one another while pursuing the same objective”. 

This kind of devil-may-care approach reminds us of China’s freelance cyberespionage outfits. These groups hack first, then shop the stolen materials around to PRC intelligence services. For these actors, more hacking means more sales opportunities.

And Vapor Panda had some success. Out of the 200 organisations targeted, CrowdStrike confirmed post-exploitation activity in nine of them. There were two “extensive intrusions” into the foreign ministries of at least two Latin American countries. A Middle East-based interior ministry was also compromised. 

Loud and fast isn’t the best strategy for all actors, however.  

On the stealthier side of the spectrum we have Silent Train, an illicit network of compromised devices used by Volt Typhoon, the group compromising US critical infrastructure in order to disrupt it in the event of a military conflict. Late last month Black Lotus Labs released a report on Silent Train. 

When it comes to operational security and stealth, the contrast between Volt Typhoon and Vapor Panda could not be more stark. Volt Typhoon has no interest in smash-and-grab intelligence gathering. Its goal is to guarantee access to American critical infrastructure at some future date and it is all about stealthy operation.

Rather than being persistent infrastructure, Silent Train is made up of short-lived chains of compromised devices that Volt Typhoon operators would route activity through to reach targets. Operators could assemble chains in under two minutes by exploiting exposed web interfaces on SOHO, IoT and other relay devices. They’d deploy just the tooling required for chained SSH access through the compromised node, including temporary SSH services, keys, configuration files and living-off-the-land utilities. Volt Typhoon removes all these artifacts when it dismantles the chains, and they can be set up and torn down in as little as ten minutes. On average they lasted four hours.

AI-assisted YOLO hacking is clearly counterproductive when sustaining access is a priority, and there is evidence that Silent Train was operated by, or at least overseen by, real people. Most of its activity occurred during human working hours, from Monday to Friday, 9am to 10pm in China Standard time. This does not appear to be an agentically driven operation. Very artisanal! How quaint!

So we have three distinct examples of how China’s cyber operators are currently operating. Vapor Panda utilises AI for speed and is unphased by noise. SilkParasite bolsters its malware arsenal with AI to muddy attribution. Volt Typhoon, the most cautious and stealthy of the three, is operating slowly and steadily, with human hands on keyboards from 9am Beijing time, Monday to Friday.

The lesson? China gonna China. But in 2026 and partially thanks to AI, it’s managed to become faster, louder or sneakier, depending on where you look.

Watch Amberleigh Jack and Tom Uren discuss this edition of the newsletter:

Three Reasons to Be Cheerful This Week:

1. AI is getting better at respecting scope: OpenAI has released its new cybersecurity model, GPT-6 Cyber, and says that it is noticeably better at the ExploitGym benchmark. Researchers have used it to find 81 new high-severity vulnerabilities in Firefox and over 900 linux privilege escalation bugs. Most importantly, OpenAI says the model was far better at staying within its authorised scope.  
2. Parts of Signal get formally verified: Google DeepMind researchers have used  a new AI framework called AlphaProof Code to formally verify that some of the Signal protocol actually works as expected. The research took just four weeks, which is very fast for this kind of work and raises hope that formal verification might be used more widely to assess other very important algorithms. That would be good, but the scope of this analysis was still quite limited and it didn’t assess the Signal app, for example. 
3. Killing secrets before they are abused: Truffle Security has launched a program to help API issuers to revoke live secrets it finds on the public internet. Truffle scans publicly accessible data such as code, packages and container images and finds lots and lots of live secrets. Issuing companies that have joined the Secrets Revocation Program will now revoke keys when alerted by Truffle. This will be a big security win because over 80 percent of the time developers do not revoke keys when alerted by email. Developers can opt out of the service. 

Sponsor Section

Tom Uren talks to Airlock Digital’s CTO and CSO Peter Baussmann about how the Essential Eight is evolving into ASD’s Essentials framework.

Risky Biz Talks

You can find the audio edition of this newsletter and other fine podcasts and interviews in the Risky Biz News feed (RSS, iTunes or Spotify).  

In our last “Between Two Nerds” discussion Tom Uren and The Grugq  talk about why PHP2Shell failed to live up to its hype. 

Or watch it on YouTube!

From Risky Bulletin:

Pro-Kremlin hacktivists dox EU politicians, law enforcement: A pro-Kremlin hacktivist group named NoName057(16) has published the personal information of 18 high-profile EU politicians and law enforcement officials.

Victims include Italian Foreign Minister Antonio Tajani, the Eurojust president Michael Schmid, Europol European Cybercrime Centre head Edvardas Šileris, German Federal Police (BKA) president Holger Münch, and Italian Postal Police chief Ivano Gabrielli.

The dox also covers three German journalists who investigated Russian hackers and influence operations—Sergej Maier, Angelique Geray, and Michael Alexander Martens.

[more on Risky Bulletin]

Some OpenSSH instances take any password longer than 256 characters: There can’t be any newsletter edition without a dumb bug, and today’s selection is a vulnerability codenamed SSHhole, discovered by Wakefield security researcher Ozgun Kultekin.

This particular beauty impacts the OpenSSH project and allows attackers to log into accounts with any random string with more than 256 characters. It doesn’t matter what the characters are, you get root as long as you type more than 256.

The bug will work against accounts that have the password disabled or use cryptographic keys to log in.

[more on Risky Bulletin]

Originally published on Risky Business News